{"id":"https://openalex.org/W7214734150","doi":"https://doi.org/10.48550/arxiv.2609.31746","title":"VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models","display_name":"VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models","publication_year":2026,"publication_date":"2026-09-23","ids":{"openalex":"https://openalex.org/W7214734150","doi":"https://doi.org/10.48550/arxiv.2609.31746"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.31746","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.31746","pdf_url":"https://arxiv.org/pdf/2609.31746","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.31746","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5051437282","display_name":"Khurram Azeem Hashmi","orcid":"https://orcid.org/0000-0003-0456-6493","observed_orcids":["https://orcid.org/0000-0003-0456-6493"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hashmi, Khurram Azeem","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045033461","display_name":"Mohammadreza Zolfaghari","orcid":"https://orcid.org/0000-0002-2973-4302","observed_orcids":["https://orcid.org/0000-0002-2973-4302"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zolfaghari, Mohammadreza","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089785034","display_name":"Changdae Park","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Changdae","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101807691","display_name":"Rishabh Jain","orcid":"https://orcid.org/0009-0000-1021-5038","observed_orcids":["https://orcid.org/0009-0000-1021-5038","https://orcid.org/0009-0001-1461-5300","https://orcid.org/0009-0008-0765-9509"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jain, Rishabh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090402362","display_name":"Nicholas Moratelli","orcid":"https://orcid.org/0000-0001-9362-5680","observed_orcids":["https://orcid.org/0000-0001-9362-5680"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Moratelli, Nicholas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101506669","display_name":"Pengfei Wei","orcid":"https://orcid.org/0000-0001-8093-0803","observed_orcids":["https://orcid.org/0000-0001-8093-0803"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Pengfei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153436377","display_name":"Louis Lu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Louis","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153490660","display_name":"Tianchi Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Tianchi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5042617322","display_name":"Amril Nazir","orcid":"https://orcid.org/0000-0001-7116-7607","observed_orcids":["https://orcid.org/0000-0001-7116-7607","https://orcid.org/0000-0003-1960-5424"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nazir, Amril","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9409000277519226,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9409000277519226,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.05079999938607216,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0027000000700354576,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/vision-language-models","display_name":"vision-language models","score":0.9559999704360962},{"id":"https://openalex.org/keywords/on-device-deployment","display_name":"on-device deployment","score":0.9620000123977661},{"id":"https://openalex.org/keywords/mobile-devices","display_name":"mobile devices","score":0.4390000104904175},{"id":"https://openalex.org/keywords/post-training-optimization","display_name":"post-training optimization","score":0.6600000262260437},{"id":"https://openalex.org/keywords/failure-mode-identification","display_name":"failure mode identification","score":0.6439999938011169},{"id":"https://openalex.org/keywords/efficiency","display_name":"efficiency","score":0.35100001096725464},{"id":"https://openalex.org/keywords/accuracy-efficiency-trade-off","display_name":"accuracy-efficiency trade-off","score":0.6700000166893005},{"id":"https://openalex.org/keywords/visual-tokenization","display_name":"visual tokenization","score":0.5659999847412109},{"id":"https://openalex.org/keywords/teacher-student-training","display_name":"teacher-student training","score":0.5989999771118164},{"id":"https://openalex.org/keywords/preference-alignment","display_name":"preference alignment","score":0.699999988079071}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.678600013256073},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.666100025177002},{"id":"https://openalex.org/C2778671685","wikidata":"https://www.wikidata.org/wiki/Q219239","display_name":"Recipe","level":2,"score":0.6437000036239624},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.5493999719619751},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5475999712944031},{"id":"https://openalex.org/C170130773","wikidata":"https://www.wikidata.org/wiki/Q216378","display_name":"Usability","level":2,"score":0.5401999950408936},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.47850000858306885},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.47780001163482666},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.46880000829696655},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43799999356269836},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.3440999984741211},{"id":"https://openalex.org/C160633673","wikidata":"https://www.wikidata.org/wiki/Q355198","display_name":"Pixel","level":2,"score":0.3314000070095062},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.31949999928474426},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.3158000111579895},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.29600000381469727},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.28679999709129333},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.28619998693466187},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2825999855995178},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.27559998631477356},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.26100000739097595},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.25540000200271606},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2547999918460846},{"id":"https://openalex.org/C2781140086","wikidata":"https://www.wikidata.org/wiki/Q557945","display_name":"Confusion","level":2,"score":0.25380000472068787}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.31746","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.31746","pdf_url":"https://arxiv.org/pdf/2609.31746","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.31746","doi":"https://doi.org/10.48550/arxiv.2609.31746","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.31746","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.31746","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.31746","pdf_url":"https://arxiv.org/pdf/2609.31746","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214734150.pdf","grobid_xml":"https://content.openalex.org/works/W7214734150.grobid-xml"},"referenced_works_count":12,"referenced_works":["https://openalex.org/W4390437694","https://openalex.org/W4404569805","https://openalex.org/W4404781880","https://openalex.org/W4407815020","https://openalex.org/W4415962969","https://openalex.org/W4416119086","https://openalex.org/W4416982472","https://openalex.org/W6947991961","https://openalex.org/W7118442321","https://openalex.org/W7151585725","https://openalex.org/W7161273375","https://openalex.org/W7172489173"],"related_works":["https://openalex.org/W4405470172","https://openalex.org/W7138903823","https://openalex.org/W4417129446","https://openalex.org/W4390437694","https://openalex.org/W7164388866","https://openalex.org/W7165564498","https://openalex.org/W7128522858","https://openalex.org/W7172448945","https://openalex.org/W7161292544","https://openalex.org/W4417099704"],"abstract_inverted_index":{"Sub-billion-parameter":[0],"Vision-Language":[1],"Models":[2],"are":[3],"increasingly":[4],"viable":[5],"for":[6],"on-device":[7,182],"deployment,":[8],"yet":[9],"compact":[10,178],"model":[11,23],"size":[12],"alone":[13],"does":[14],"not":[15],"guarantee":[16],"usability.":[17,183],"On":[18],"a":[19,22,70,76,121,162],"phone,":[20],"such":[21],"can":[24],"still":[25],"require":[26],"more":[27],"than":[28],"two":[29,106,112],"minutes":[30],"to":[31,56,64,167],"produce":[32],"its":[33,137],"first":[34],"token.":[35],"On-device":[36],"usability":[37],"depends":[38],"on":[39,161],"three":[40,174],"axes:":[41],"accuracy,":[42],"efficiency,":[43],"and":[44,60,88,94,142],"behavioral":[45,117],"reliability;":[46],"standard":[47],"benchmarks":[48],"miss":[49],"the":[50,80,109,128,151],"third,":[51],"with":[52,101,105,115,145],"answers":[53],"too":[54,62],"short":[55],"expose":[57],"doom":[58],"loops":[59],"prompts":[61],"benign":[63],"probe":[65],"adversarial":[66],"safety.":[67],"We":[68],"introduce":[69],"diagnosis-driven":[71],"post-training":[72],"recipe":[73,110],"in":[74],"which":[75],"teacher":[77],"VLM":[78],"stress-tests":[79],"student,":[81],"uncovers":[82],"failure":[83],"modes":[84],"beyond":[85],"human":[86],"priors,":[87],"converts":[89],"them":[90],"into":[91],"targeted":[92],"supervision":[93],"preference":[95],"alignment,":[96],"supplementing":[97],"generic":[98],"data":[99],"scaling":[100],"failure-driven":[102],"optimization.":[103],"Coupled":[104],"visual-token":[107],"policies,":[108],"yields":[111],"accuracy-efficiency":[113],"variants":[114],"improved":[116],"reliability.":[118],"\\textbf{\\NanoFull}":[119],"attains":[120],"62.3":[122],"normalized":[123],"average":[124],"over":[125,136],"17":[126],"benchmarks,":[127],"highest":[129],"among":[130],"openly":[131],"released":[132],"$\\sim$0.5B":[133],"models,":[134],"+7.4":[135],"base":[138],"at":[139,148],"identical":[140],"architecture":[141],"token":[143],"budget,":[144],"doom-loop":[146],"rates":[147],"or":[149],"below":[150],"strongest":[152],"baseline's.":[153],"\\textbf{\\FlashFull}":[154],"retains":[155],"61.4":[156],"while":[157],"cutting":[158],"warm":[159],"time-to-first-token":[160],"Pixel":[163],"9":[164],"from":[165],"138\\,s":[166],"6.1\\,s":[168],"(23$\\times$).":[169],"By":[170],"jointly":[171],"addressing":[172],"all":[173],"axes,":[175],"we":[176],"move":[177],"VLMs":[179],"toward":[180],"practical":[181]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-11-11T12:16:26.678958","created_date":"2026-09-30T00:00:00"}